To illustrate, consider the curiously old-fashioned approach companies typically take to upgrading their legacy systems. It starts when something old stops working. Perhaps an aging mainframe fails often, resulting in seemingly never-ending maintenance costs, or an outage destroys transaction data.